Proceedings of the Annual Meeting of the Cognitive Science Society

Children's early vocabulary is often skewed towards nouns and away from verbs, but these biases are thought to vary cross-linguistically. Prior work attempting to explore this variation has often been limited in language coverage, predictors, or bias estimation methods. Drawing on large-scale cross-linguistic resources (Wordbank, WALS, GramBank, and CHILDES), we provide a systematic decomposition of early lexical category biases into typological, frequency-based, and cultural components. We first derive noun, verb, and adjective bias estimates from early productive vocabularies across 32 languages and dialects. We then test whether cross-linguistic variation in these biases can be predicted by grammatical features and part-of-speech frequency in child-directed speech, as well as broader cultural factors. Several of these predictors reliably explain variation in noun and verb bias, whereas adjective bias is comparatively poorly predicted. These findings highlight multiple classes of factors shaping variation in early lexical biases.